Files
light-petal-recognition/scripts/prepare_dataset.py
T

131 lines
4.3 KiB
Python

"""
数据集下载和准备脚本
"""
import os
import argparse
import shutil
from pathlib import Path
def create_dummy_dataset(output_dir, num_classes=10, images_per_class=100):
"""
创建示例数据集结构
Args:
output_dir: 输出目录
num_classes: 类别数量
images_per_class: 每个类别的图像数量
"""
print(f"创建数据集结构: {output_dir}")
# 创建训练集和验证集
for split in ['train', 'val']:
split_dir = os.path.join(output_dir, split)
os.makedirs(split_dir, exist_ok=True)
# 创建类别目录
for class_id in range(num_classes):
class_dir = os.path.join(split_dir, f'class_{class_id}')
os.makedirs(class_dir, exist_ok=True)
# 创建占位符文件
readme_path = os.path.join(class_dir, 'README.txt')
with open(readme_path, 'w', encoding='utf-8') as f:
f.write(f'类别 {class_id}: {class_id}个光瓣\n')
f.write(f'请在此目录放置 {images_per_class} 张图像\n')
f.write('支持的格式: .jpg, .jpeg, .png\n')
print("数据集结构创建完成!")
print("\n目录结构:")
print(f"{output_dir}/")
print(" ├── train/")
print(" │ ├── class_0/ (0个光瓣)")
print(" │ ├── class_1/ (1个光瓣)")
print(" │ └── ...")
print(" └── val/")
print(" ├── class_0/")
print(" └── ...")
print("\n请将图像放入对应的类别目录中")
def split_dataset(source_dir, output_dir, train_ratio=0.8):
"""
将数据集分割为训练集和验证集
Args:
source_dir: 源数据目录
output_dir: 输出目录
train_ratio: 训练集比例
"""
import random
print(f"分割数据集: {source_dir} -> {output_dir}")
# 创建输出目录
train_dir = os.path.join(output_dir, 'train')
val_dir = os.path.join(output_dir, 'val')
os.makedirs(train_dir, exist_ok=True)
os.makedirs(val_dir, exist_ok=True)
# 遍历类别
for class_name in os.listdir(source_dir):
class_path = os.path.join(source_dir, class_name)
if not os.path.isdir(class_path):
continue
# 获取所有图像
images = [f for f in os.listdir(class_path)
if f.lower().endswith(('.png', '.jpg', '.jpeg'))]
# 随机打乱
random.shuffle(images)
# 分割
split_idx = int(len(images) * train_ratio)
train_images = images[:split_idx]
val_images = images[split_idx:]
# 创建类别目录
train_class_dir = os.path.join(train_dir, class_name)
val_class_dir = os.path.join(val_dir, class_name)
os.makedirs(train_class_dir, exist_ok=True)
os.makedirs(val_class_dir, exist_ok=True)
# 复制文件
for img in train_images:
src = os.path.join(class_path, img)
dst = os.path.join(train_class_dir, img)
shutil.copy2(src, dst)
for img in val_images:
src = os.path.join(class_path, img)
dst = os.path.join(val_class_dir, img)
shutil.copy2(src, dst)
print(f"{class_name}: {len(train_images)} 训练, {len(val_images)} 验证")
print("数据集分割完成!")
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='数据集准备工具')
parser.add_argument('--mode', type=str, choices=['create', 'split'],
required=True, help='操作模式')
parser.add_argument('--output', type=str, required=True,
help='输出目录')
parser.add_argument('--source', type=str, default=None,
help='源数据目录(split模式)')
parser.add_argument('--num_classes', type=int, default=10,
help='类别数量(create模式)')
parser.add_argument('--train_ratio', type=float, default=0.8,
help='训练集比例(split模式)')
args = parser.parse_args()
if args.mode == 'create':
create_dummy_dataset(args.output, args.num_classes)
elif args.mode == 'split':
if args.source is None:
print("错误: split模式需要指定--source参数")
else:
split_dataset(args.source, args.output, args.train_ratio)